# Changes in `2.6.1` <a name="changelog-pytorch"></a> ## PyTorch Lightning <details open><summary>Added</summary> - Added method chaining support to `LightningModule.freeze()` and `LightningModule.unfreeze()` by returning `self` ([#21469](https://github.com/Lightning-AI/pytorch-lightning/pull/21469)) - Added litlogger integration([#21430](https://github.com/Lightning-AI/pytorch-lightning/pull/21430)) <details open><summary>Deprecated</summary> - Deprecated `to_torchscript` method due to deprecation of TorchScript in PyTorch ([#21397](https://github.com/Lightning-AI/pytorch-lightning/pull/21397)) </details> <details open><summary>Removed</summary> - Removed support for Python 3.9 due to end-of-life status ([#21398](https://github.com/Lightning-AI/pytorch-lightning/pull/21398)) </details> <details open><summary>Fixed</summary> - Fixed `save_hyperparameters(ignore=...)` behavior so subclass ignore rules override base class rules (#[21490](https://github.com/Lightning-AI/pytorch-lightning/pull/21490)) - Fixed `LightningDataModule.load_from_checkpoint` to restore the datamodule subclass and hyperparameters ([#21478](https://github.com/Lightning-AI/pytorch-lightning/pull/21478)) - Fixed ``ModelParallelStrategy`` single-file checkpointing when ``torch.compile`` wraps the model so optimizer states no longer raise ``KeyError`` during save ([#21357](https://github.com/Lightning-AI/pytorch-lightning/issues/21357)) - Sanitize profiler filenames when saving to avoid crashes due to invalid characters ([#21395](https://github.com/Lightning-AI/pytorch-lightning/pull/21395)) - Fixed `StochasticWeightAveraging` with infinite epochs ([#21396](https://github.com/Lightning-AI/pytorch-lightning/pull/21396)) - Fixed `_generate_seed_sequence_sampling` function not producing unique seeds ([#21399](https://github.com/Lightning-AI/pytorch-lightning/pull/21399)) - Fixed `ThroughputMonitor` callback emitting warnings too frequently ([#21453](https://github.com/Lightning-AI/pytorch-lightning/pull/21453)) </details> ## Lightning Fabric <details open><summary>Added</summary> - Exposed `weights_only` argument for loading checkpoints in `Fabric.load()` and `Fabric.load_raw()` ([#21470](https://github.com/Lightning-AI/pytorch-lightning/pull/21470)) </details> <details open><summary>Fixed</summary> - Fixed `DistributedSamplerWrapper` not forwarding `set_epoch` to the underlying sampler ([#21454](https://github.com/Lightning-AI/pytorch-lightning/pull/21454)) - Fixed DDP notebook CUDA fork check to allow passive initialization when CUDA is not actively used ([#21402](https://github.com/Lightning-AI/pytorch-lightning/pull/21402)) </details> </br> **Full commit list**: [2.6.0 -> 2.6.1](https://github.com/Lightning-AI/pytorch-lightning/compare/2.6.0...2.6.1) <a name="contributors"></a> # Contributors **New Contributors** * @arrdel made their first contribution in https://github.com/Lightning-AI/pytorch-lightning/pull/21402 * @CodeVishal-17 made their first contribution in https://github.com/Lightning-AI/pytorch-lightning/pull/21470 * @aditya0by0 made their first contribution in https://github.com/Lightning-AI/pytorch-lightning/pull/21478 We thank **all folks** who submitted issues, features, fixes and doc changes. It's the only way we can **collectively** make Lightning :zap: better for everyone, nice job! In particular, we would like to thank the authors of the pull-requests above

# Changes in `2.6.0` <a name="changelog-pytorch"></a> ## PyTorch Lightning <details open><summary>Added</summary> - Added `WeightAveraging` callback that wraps the PyTorch `AveragedModel` class ([#20545](https://github.com/Lightning-AI/pytorch-lightning/pull/20545)) - Added Torch-Tensorrt integration with `LightningModule` ([#20808](https://github.com/Lightning-AI/pytorch-lightning/pull/20808)) - Added time-based validation support though `val_check_interval` ([#21071](https://github.com/Lightning-AI/pytorch-lightning/pull/21071)) - Added attributes to access stopping reason in `EarlyStopping` callback ([#21188](https://github.com/Lightning-AI/pytorch-lightning/pull/21188)) - Added support for variable batch size in `ThroughputMonitor` ([#20236](https://github.com/Lightning-AI/pytorch-lightning/pull/20236)) - Added `EMAWeightAveraging` callback that wraps Lightning's `WeightAveraging` class ([#21260](https://github.com/Lightning-AI/pytorch-lightning/pull/21260)) </details> <details open><summary>Changed</summary> - Expose `weights_only` argument for `Trainer.{fit,validate,test,predict}` and let `torch` handle default value ([#21072](https://github.com/Lightning-AI/pytorch-lightning/pull/21072)) - Default to `RichProgressBar` and `RichModelSummary` if the rich package is available. Fallback to TQDMProgressBar and ModelSummary otherwise ([#20896](https://github.com/Lightning-AI/pytorch-lightning/pull/20896)) - Add MPS accelerator support for mixed precision ([#21209](https://github.com/Lightning-AI/pytorch-lightning/pull/21209)) </details> <details open><summary>Fixed</summary> - Fixed edgecase when `max_trials` is reached in `Tuner.scale_batch_size` ([#21187](https://github.com/Lightning-AI/pytorch-lightning/pull/21187)) - Fixed case where `LightningCLI` could not be initialized with `trainer_default` containing callbacks ([#21192](https://github.com/Lightning-AI/pytorch-lightning/pull/21192)) - Fixed missing reset when `ModelPruning` is applied with lottery ticket hypothesis ([#21191](https://github.com/Lightning-AI/pytorch-lightning/pull/21191)) - Fixed preventing recursive symlink creation iwhen `save_last='link'` and `save_top_k=-1` ([#21186](https://github.com/Lightning-AI/pytorch-lightning/pull/21186)) - Fixed `last.ckpt` being created and not linked to another checkpoint ([#21244](https://github.com/Lightning-AI/pytorch-lightning/pull/21244)) - Fixed bug that prevented `BackboneFinetuning` from being used together with `LearningRateFinder` ([#21224](https://github.com/Lightning-AI/pytorch-lightning/pull/21224)) - Fixed `ModelPruning` sparsity logging bug that caused incorrect sparsity percentages ([#21223](https://github.com/Lightning-AI/pytorch-lightning/pull/21223)) - Fixed `LightningCLI` loading of hyperparameters from `ckpt_path` failing for subclass model mode ([#21246](https://github.com/Lightning-AI/pytorch-lightning/pull/21246)) - Fixed check the init args only when the given frames are in `__init__` method ([#21227](https://github.com/Lightning-AI/pytorch-lightning/pull/21227)) - Fixed how `ThroughputMonitor` calculated training time ([#21291](https://github.com/Lightning-AI/pytorch-lightning/pull/21291)) - Fixed synchronization of gradients in manual optimization with `DDPStrategy(static_graph=True)` ([#21251](https://github.com/Lightning-AI/pytorch-lightning/pull/21251)) - Fixed FSDP mixed precision semantics and added user warning ([#21361](https://github.com/Lightning-AI/pytorch-lightning/pull/21361)) </details> ## Lightning Fabric <details open><summary>Changed</summary> - Expose `weights_only` argument for `Trainer.{fit,validate,test,predict}` and let `torch` handle default value ([#21072](https://github.com/Lightning-AI/pytorch-lightning/pull/21072)) - Set `_DeviceDtypeModuleMixin._device` from torch's default device function ([#21164](https://github.com/Lightning-AI/pytorch-lightning/pull/21164)) - Added kwargs-filtering for `Fabric.call` to support different callback method signatures ([#21258](https://github.com/Lightning-AI/pytorch-lightning/pull/21258)) </details> <details open><summary>Fixed</summary> - Fixed issue in detecting MPIEnvironment with partial mpi4py installation ([#21353](https://github.com/Lightning-AI/pytorch-lightning/pull/21353)) - Learning rate scheduler is stepped at the end of epoch when `on_train_batch_start` returns -1 ([#21296](https://github.com/Lightning-AI/pytorch-lightning/issues/21296)). - Fixed FSDP mixed precision semantics and added user warning ([#21361](https://github.com/Lightning-AI/pytorch-lightning/pull/21361)) </details> </br> **Full commit list**: [2.5.4 -> 2.5.5](https://github.com/Lightning-AI/pytorch-lightning/compare/2.5.6...2.6.0) <a name="contributors"></a> # Contributors We thank **all folks** who submitted issues, features, fixes and doc changes. It's the only way we can **collectively** make Lightning :zap: better for everyone, nice job! In particular, we would like to thank the authors of the pull-requests above

# Changes in `2.5.6` <a name="changelog-pytorch"></a> ## PyTorch Lightning <details open><summary>Changed</summary> - Add `name()` function to accelerator interface (([#21325](https://github.com/Lightning-AI/pytorch-lightning/pull/21325))) </details> <details open><summary>Removed</summary> - Remove support for deprecated and archived lightning-habana package ([#21327](https://github.com/Lightning-AI/pytorch-lightning/pull/21327)) </details>

# Changes in `2.5.5` <a name="changelog-pytorch"></a> ## PyTorch Lightning <details open><summary>Changed</summary> - Include `exclude_frozen_parameters` to `DeepSpeedStrategy` ([#21060](https://github.com/Lightning-AI/pytorch-lightning/pull/21060)) - Include `PossibleUserWarning` that is raised if modules are in eval mode when training starts ([#21146](https://github.com/Lightning-AI/pytorch-lightning/pull/21146)) </details> <details open><summary>Fixed</summary> - Fixed `LightningCLI` not using `ckpt_path` hyperparameters to instantiate classes ([#21116](https://github.com/Lightning-AI/pytorch-lightning/pull/21116)) - Fixed callbacks by defer step/time-triggered `ModelCheckpoint` saves until validation metrics are available ([#21106](https://github.com/Lightning-AI/pytorch-lightning/pull/21106)) - Fixed with adding a missing device id for pytorch 2.8 ([#21105](https://github.com/Lightning-AI/pytorch-lightning/pull/21105)) - Fixed `TQDMProgressBar` not resetting correctly when using both a finite and iterable dataloader ([#21147](https://github.com/Lightning-AI/pytorch-lightning/pull/21147)) - Fixed cleanup of temporary files from `Tuner` on crashes ([#21162](https://github.com/Lightning-AI/pytorch-lightning/pull/21162)) </details> <a name="changelog-fabric"></a> ## Lightning Fabric <details open><summary>Changed</summary> - Include `exclude_frozen_parameters` to `DeepSpeedStrategy` ([#21060](https://github.com/Lightning-AI/pytorch-lightning/pull/21060)) - Let `_get_default_process_group_backend_for_device` support more hardware platforms ( [#21057](https://github.com/Lightning-AI/pytorch-lightning/pull/21057), [#21093](https://github.com/Lightning-AI/pytorch-lightning/pull/21093)) </details> <details open><summary>Fixed</summary> - Fixed with adding a missing device id for pytorch 2.8 ([#21105](https://github.com/Lightning-AI/pytorch-lightning/pull/21105)) - Respecting `verbose=False` in `seed_everything` when no seed is provided ([#21161](https://github.com/Lightning-AI/pytorch-lightning/pull/21161)) </details> </br> **Full commit list**: [2.5.4 -> 2.5.5](https://github.com/Lightning-AI/pytorch-lightning/compare/2.5.4...2.5.5) <a name="contributors"></a> # Contributors We thank **all folks** who submitted issues, features, fixes and doc changes. It's the only way we can **collectively** make Lightning :zap: better for everyone, nice job! In particular, we would like to thank the authors of the pull-requests above, in no particular order: @Borda, @KAVYANSHTYAGI, @littlebullGit, @mauvilsa, @SkafteNicki, @taozhiwei Thank you :heart: and we hope you'll keep them coming!

# Changes in `2.5.4` <a name="changelog-pytorch"></a> ## PyTorch Lightning <details open><summary>Fixed</summary> - Fixed `AsyncCheckpointIO` snapshots tensors to avoid race with parameter mutation ([#21079](https://github.com/Lightning-AI/pytorch-lightning/pull/21079)) - Fixed `AsyncCheckpointIO` threadpool exception if calling fit or validate more than one ([#20952](https://github.com/Lightning-AI/pytorch-lightning/pull/20952)) - Fixed learning rate not being correctly set after using `LearningRateFinder` callback ([#21068](https://github.com/Lightning-AI/pytorch-lightning/pull/21068)) - Fixed misalignment column while using rich model summary in `DeepSpeedstrategy` ([#21100](https://github.com/Lightning-AI/pytorch-lightning/pull/21100)) - Fixed `RichProgressBar` crashing when sanity checking using val dataloader with 0 len ([#21108](https://github.com/Lightning-AI/pytorch-lightning/pull/21108)) </details> <a name="changelog-fabric"></a> ## Lightning Fabric <details open><summary>Changed</summary> - Added support for NVIDIA H200 GPUs in `get_available_flops` ([#20913](https://github.com/Lightning-AI/pytorch-lightning/pull/21119)) </details> </br> **Full commit list**: [2.5.3 -> 2.5.4](https://github.com/Lightning-AI/pytorch-lightning/compare/2.5.3...2.5.4) <a name="contributors"></a> # Contributors We thank **all folks** who submitted issues, features, fixes and doc changes. It's the only way we can **collectively** make Lightning :zap: better for everyone, nice job! In particular, we would like to thank the authors of the pull-requests above, in no particular order: @fnhirwa, @GdoongMathew, @jjh42, @littlebullGit, @SkafteNicki Thank you :heart: and we hope you'll keep them coming!

# Notable changes in this release <a name="changelog-pytorch"></a> ## PyTorch Lightning <details open><summary>Changed</summary> - Add `enable_autolog_hparams` argument to Trainer ([#20593](https://github.com/Lightning-AI/pytorch-lightning/pull/20593)) - Add `toggled_optimizer(optimizer)` method to the LightningModule, which is a context manager version of `toggle_optimize` and `untoggle_optimizer` ([#20771](https://github.com/Lightning-AI/pytorch-lightning/pull/20771)) - For cross-device local checkpoints, instruct users to install `fsspec>=2025.5.0` if unavailable ([#20780](https://github.com/Lightning-AI/pytorch-lightning/pull/20780)) - Check param is of `nn.Parameter` type for pruning sanitization ([#20783](https://github.com/Lightning-AI/pytorch-lightning/pull/20783)) </details> <details open><summary>Fixed</summary> - Fixed `save_hyperparameters` not working correctly with `LightningCLI` when there are parsing links applied on instantiation ([#20777](https://github.com/Lightning-AI/pytorch-lightning/pull/20777)) - Fixed `logger_connector` has an edge case where step can be a float ([#20692](https://github.com/Lightning-AI/pytorch-lightning/pull/20692)) - Fixed Synchronize SIGTERM Handling in DDP to Prevent Deadlocks ([#20825](https://github.com/Lightning-AI/pytorch-lightning/pull/20825)) - Fixed case-sensitive model name ([#20661](https://github.com/Lightning-AI/pytorch-lightning/pull/20661)) - CLI: resolve jsonargparse deprecation warning ([#20802](https://github.com/Lightning-AI/pytorch-lightning/pull/20802)) - Fix: move `check_inputs` to the target device if available during `to_torchscript` ([#20873](https://github.com/Lightning-AI/pytorch-lightning/pull/20873)) - Fixed progress bar display to correctly handle iterable dataset and `max_steps` during training ([#20869](https://github.com/Lightning-AI/pytorch-lightning/pull/20869)) - Fixed problem for silently supporting `jsonnet` ([#20899](https://github.com/Lightning-AI/pytorch-lightning/pull/20899)) </details> <a name="changelog-fabric"></a> ## Lightning Fabric <details open><summary>Changed</summary> - Ensure correct device is used for autocast when mps is selected as Fabric accelerator ([#20876](https://github.com/Lightning-AI/pytorch-lightning/pull/20876)) </details> <details open><summary>Removed</summary> - Fix: `TransformerEnginePrecision` conversion for layers with `bias=False` ([#20805](https://github.com/Lightning-AI/pytorch-lightning/pull/20805)) </details> </br> **Full commit list**: [2.5.1 -> 2.5.2](https://github.com/Lightning-AI/pytorch-lightning/compare/2.5.1...2.5.2) <a name="contributors"></a> # Contributors We thank **all folks** who submitted issues, features, fixes, and doc changes. It's the only way we can **collectively** make Lightning :zap: better for everyone, nice job! In particular, we would like to thank the authors of the pull-requests above, in no particular order: @adamjstewart, @Armannas, @bandpooja, @Borda, @chanokin, @duydl, @GdoongMathew, @KAVYANSHTYAGI, @mauvilsa, @muthissar, @rustamzh, @siemdejong Thank you :heart: and we hope you'll keep them coming!

**Full Changelog**: https://github.com/Lightning-AI/pytorch-lightning/compare/2.5.1...2.5.1.post0

# Changes <a name="changelog-pytorch"></a> ## PyTorch Lightning <details open><summary>Changed</summary> - Allow LightningCLI to use a customized argument parser class ([#20596](https://github.com/Lightning-AI/pytorch-lightning/pull/20596)) - Change `wandb` default x-axis to `tensorboard`'s `global_step` when `sync_tensorboard=True` ([#20611](https://github.com/Lightning-AI/pytorch-lightning/pull/20611)) - Added a new `checkpoint_path_prefix` parameter to the MLflow logger which can control the path to where the MLflow artifacts for the model checkpoints are stored ([#20538](https://github.com/Lightning-AI/pytorch-lightning/pull/20538)) - CometML logger was updated to support the recent Comet SDK ([#20275](https://github.com/Lightning-AI/pytorch-lightning/pull/20275)) - bump: testing with latest `torch` 2.6 ([#20509](https://github.com/Lightning-AI/pytorch-lightning/pull/20509)) </details> <details open><summary>Fixed</summary> - Fixed CSVLogger logging hyperparameter at every write which increases latency ([#20594](https://github.com/Lightning-AI/pytorch-lightning/pull/20594)) - Fixed OverflowError when resuming from checkpoint with an iterable dataset ([#20565](https://github.com/Lightning-AI/pytorch-lightning/issues/20565)) - Fixed swapped `_R_co` and `_P` to prevent type error ([#20508](https://github.com/Lightning-AI/pytorch-lightning/issues/20508)) - Always call `WandbLogger.experiment` first in `_call_setup_hook` to ensure `tensorboard` logs can sync to `wandb` ([#20610](https://github.com/Lightning-AI/pytorch-lightning/pull/20610)) - Fixed TBPTT example ([#20528](https://github.com/Lightning-AI/pytorch-lightning/pull/20528)) - Fixed test compatibility as AdamW became a subclass of Adam ([#20574](https://github.com/Lightning-AI/pytorch-lightning/pull/20574)) - Fixed file extension of model checkpoints uploaded by NeptuneLogger ([#20581](https://github.com/Lightning-AI/pytorch-lightning/pull/20581)) - Reset trainer variable `should_stop` when `fit` is called ([#19177](https://github.com/Lightning-AI/pytorch-lightning/pull/19177)) - Fixed making `WandbLogger` upload models from all `ModelCheckpoint` callbacks, not just one ([#20191](https://github.com/Lightning-AI/pytorch-lightning/pull/20191)) - Error when logging to MLFlow deleted experiment ([#20556](https://github.com/Lightning-AI/pytorch-lightning/pull/20556)) </details> <a name="changelog-fabric"></a> ## Lightning Fabric <details open><summary>Changed</summary> - Added logging support for a list of dicts without collapsing to a single key ([#19957](https://github.com/Lightning-AI/pytorch-lightning/issues/19957)) - bump: testing with latest `torch` 2.6 ([#20509](https://github.com/Lightning-AI/pytorch-lightning/pull/20509)) </details> <details open><summary>Removed</summary> - Removed legacy support for `lightning run model`; use `fabric run` instead. ([#20588](https://github.com/Lightning-AI/pytorch-lightning/pull/20588)) </details> </br> **Full commit list**: [2.5.0 -> 2.5.1](https://github.com/Lightning-AI/pytorch-lightning/compare/2.5.0...2.5.1) <a name="contributors"></a> # Contributors We thank **all folks** who submitted issues, features, fixes and doc changes. It's the only way we can **collectively** make Lightning :zap: better for everyone, nice job! In particular, we would like to thank the authors of the pull-requests above, in no particular order: @benglewis, @Borda, @cgebbe, @duydl, @haifeng-jin, @japdubengsub, @justusschock, @lantiga, @mauvilsa, @millskyle, @ringohoffman, @ryan597, @senarvi, @TresYap Thank you :heart: and we hope you'll keep them coming!